A system and method is disclosed for managing data in a non-volatile memory. The system may include a non-volatile memory having multiple non-volatile memory sub-drives. A controller of the memory system is configured to route incoming host data to a desired sub-drive, keep data within the same sub-drive as its source during a garbage collection operation, and re-map data between sub-drives, separate from any garbage collection operation, when a sub-drive overflows its designated amount logical address space. The method may include initial data sorting of host writes into sub-drives based on any number of hot/cold sorting functions. In one implementation, the initial host write data sorting may be based on a host list of recently written blocks for each sub-drive and a second write to a logical address encompassed by the list may trigger routing the host write to a hotter sub-drive than the current sub-drive.
|
8. A non-volatile memory system comprising:
a non-volatile memory having a plurality of sub-drives;
a controller in communication with the plurality of sub-drives, the controller configured to:
sort data associated with a host write command, as the data associated with the host write command is received from a host, into one of the plurality of sub-drives based on a determined data temperature of the data associated with the host write command; and
sort data already stored in a first sub-drive of the plurality of sub-drives into a different sub-drive of the plurality of sub-drives by logically remapping a portion of the data already stored in the first sub-drive, without rewriting the portion of the data into a different physical location, independently of any garbage collection operation in the first sub-drive.
14. A method for managing data in a memory system having a controller in communication with a non-volatile memory having a plurality of sub-drives, the method comprising the controller:
receiving a host data write at the memory system;
directing the host data write to one of the plurality of sub-drives based on a first sorting technique;
determining whether an amount of valid data in one of the plurality of sub-drives exceeds a predetermined amount of a logical address space for the one of the plurality of sub-drives; and
when the amount of valid data in the one of the plurality of sub-drives exceeds the predetermined amount, re-mapping a superblock from the one of the plurality of sub-drives to another of the plurality of sub-drives, without rewriting any data from the superblock to another physical location, based on a current data temperature of the superblock.
1. A method for managing data in a memory system having a controller in communication with a non-volatile memory having a plurality of sub-drives, the method comprising the controller:
receiving a host data write at the memory system;
directing the host data write to one of the plurality of sub-drives based on a first sorting technique;
initiating a garbage collection operation in a particular sub-drive in response to a detected garbage collection trigger;
moving valid data during the garbage collection operation from a source superblock in the particular sub-drive only to a relocation superblock in the particular sub-drive such that the valid data remains in the particular sub-drive;
determining whether a proportion of a total logical address space of the non-volatile memory currently associated with one of the plurality of sub-drives exceeds a predetermined threshold; and
when the proportion exceeds the predetermined threshold for the one of the plurality of sub-drives, re-mapping a superblock from the one of the plurality of sub-drives to another of the plurality of sub-drives, based on a current data temperature of the superblock, independently of any garbage collection operation.
2. The method of
3. The method of
4. The method of
first selecting a sub-drive from which to select the source superblock based on an analysis of write amplification for an entirety of the non-volatile memory for a current workload.
5. The method of
6. The method of
7. The method of
9. The non-volatile memory system of
10. The non-volatile memory system of
select one of the plurality of sub-drives as the first sub-drive from which to logically remap the portion of the data in response to an amount of valid data in the one of the plurality of sub-drives exceeding a predetermined amount of logical address space assigned to the one of the plurality of sub-drives.
11. The non-volatile memory system of
the portion of the data already stored in the first sub-drive comprises a superblock of the first sub-drive;
the different sub-drive comprises a next colder sub-drive of the plurality of sub-drives, wherein the next colder sub-drive comprises a sub-drive associated with data having a data temperature less than a data temperature of data associated with the first sub-drive; and
to remap the portion of the data already stored in the first sub-drive, the controller is further configured to select a coldest superblock of the first sub-drive and logically remap the coldest superblock to the different sub-drive.
12. The non-volatile memory system of
the portion of the data already stored in the first sub-drive comprises a superblock of the first sub-drive;
the different sub-drive comprises a next hotter sub-drive of the plurality of sub-drives, wherein the next hotter sub-drive comprises a sub-drive associated with data having a data temperature greater than a data temperature of data associated with the first sub-drive; and
to remap the portion of the data already stored in the first sub-drive, the controller is further configured to select a hottest superblock of the first sub-drive and logically remap the hottest superblock to the different sub-drive.
13. The non-volatile memory system of
a free block pool, the free block pool comprising a plurality of superblocks in the non-volatile memory assignable to any of the plurality of sub-drives for data storage; and
wherein the controller is further configured to:
initiate a garbage collection operation in one of the plurality of sub-drives in response to detecting that an amount of superblocks in the free block pool has fallen below a predetermined minimum threshold; and
select a source superblock for the garbage collection operation from a sub-drive having an amount of overprovisioning that is greater than a target overprovisioning for the sub-drive.
15. The method of
16. The method of
remapping the superblock to a next colder sub-drive of the plurality of sub-drives, wherein the next colder sub-drive comprises a sub-drive associated with data having a data temperature less than a data temperature of data associated with the one of the plurality of sub-drives.
17. The method of
18. The method of
remapping the superblock to a next hotter sub-drive, the next hotter sub-drive comprising a sub-drive associated with data having a data temperature greater than a data temperature of data associated with the one of the plurality of sub-drives.
19. The method of
20. The method of
writing the host data, the host data associated with logical block addresses, to a next hotter sub-drive relative to last sub-drive that previous data associated with the logical block addresses was written to.
|
This application claims the benefit of U.S. Provisional App. Ser. No. 62/518,513, filed Jun. 12, 2017, pending, the entirety of which is hereby incorporated herein by reference.
Storage systems, such as solid state drives (SSDs) including NAND flash memory, are commonly used in electronic systems ranging from consumer products to enterprise-level computer systems. SSDs and similar storage devices utilizing block-oriented architectures share a common issue: the need to create space for writing new data by collecting sparsely distributed data into a smaller number of blocks. This process is referred to as “garbage collection”. The need for garbage collection in many block-oriented storage devices is generally due to the inability to write in place to memory, and the mismatch between write granularity and erase granularity in those storage devices.
The garbage collection process may introduce a significant burden on processing resources which, in turn, may reduce SSD performance. Garbage collection involves reading valid data from a block of non-volatile memory that is to be reused and writing it back to a new block. Many real-life data workloads, notably except uniform random and sequential, have different write densities for different logical areas, with some data being ‘hot’ or frequently written, and ‘cold’ or less frequently written. When data of different temperatures is mixed together, the SSD can experience significant write amplification, where write amplification refers to the physical amount of data written or copied above the logical amount of data received.
In order to address write amplification issues, and to reduce data management overhead burdens that can be created in addressing write amplification issues, a system and method for managing data in a non-volatile memory (NVM) system having multiple sub-drives is disclosed below.
As described herein, write amplification issues that may occur in a legacy NVM system due to the mixture of different data groups, such as different temperature data, may be addressed at several different points in the operation of the NVM system. For example, in one implementation the method and system separates hot/cold data sorting from garbage collection for data that has been already been written to a sub-drive. In another implementation that may be used in combination with, or independently from the separated garbage collection and hot/cold data sorting noted above, the system and method may include an initial sort of data into a desired sub-drive as it is being written from the host.
Thus, aspects that may be usable separately or in combination may include:
Direct Host Write Path
The NVM system may be configured to strive for an optimal direct write data path to different sub-drives, where the host writes are directly routed to separate sub-drives assigned for specific data groups (e.g., ‘hot’ vs. ‘cold’) as the intended the final destination. The sub-drive destination decision may be done ‘on-the-fly’, as the data is written by the host, to obtain a direct write and avoid any penalty of an intermediate write that may require a subsequent sort. The host write of a logical block address (LBA) or logical group may be made to any sub-drive.
Local Garbage Collection
Each sub-drive may have its own independent garbage collection so that no data would move from one sub-drive to another unless resizing is required. This may avoid problems where data would otherwise be re-classified as ‘colder’ and get moved to another sub-drive just because other data in the block became obsolete and the block was selected for reclaim (garbage collection) operation. As described herein, the new free block generated by the garbage collection operation would go into a shared free block pool and may be reused by either the same sub-drive or a different sub-drive. In different embodiments, the size of a sub-drive may either be fixed or change according to a re-sizing algorithm if the workload changes.
Passive Super-Block Re-Assignment.
If sub-drive re-sizing is triggered, it may be accomplished by making sub-drives smaller via re-assigning the ‘coldest’ super-block to a ‘colder’ sub-drive, for example, the next ‘colder’ sub-drive. The ‘coldest’ super-blocks may be moved between sub-drives without physically copying data from one sub-drive to another to avoid a write amplification penalty. As described below, the least recently written block may be identified as containing the ‘coldest’ data in the particular sub-drive. For each sub-drive, the write order of superblocks by the host may be maintained as a basis to determine the temperature of the superblocks.
Overprovisioning Management
As described in greater detail below, the overprovisioning (OP) targets for each sub-drive may be based on a workload analysis for each sub-drive. The workload analysis may be used for calculations of OP, sub-drive sizes and/or data routing and may be based on tracking hit rate per sub-drive, based on the current data location. The target OP of each sub-drive may be determined from the workload analysis and then be used to determine from which sub-drive a source superblock should be selected for garbage collection.
According to a first aspect, a method for managing data in a memory system is disclosed. The method may be executed in a non-volatile memory system having a controller in communication with a non-volatile memory having a plurality of sub-drives. The method may include the controller receiving a host data write at the memory system and directing the host data write to one of the plurality of sub-drives based on a first sorting technique. The method may also include initiating a garbage collection operation in a particular sub-drive in response to a detected garbage collection trigger and moving valid data during the garbage collection operation from a source superblock in the particular sub-drive only to a relocation superblock in the particular sub-drive such that the valid data remains in the particular sub-drive. Additionally, the method may include determining whether a proportion of a total logical address space of the non-volatile memory currently associated with one of the plurality of sub-drives exceeds a predetermined threshold. When the proportion exceeds the predetermined threshold for the one of the plurality of sub-drives, the method includes re-mapping a superblock from the one of the plurality of sub-drives to another of the plurality of sub-drives based on a current data temperature of the superblock independently of any garbage collection operation.
According to another aspect, a non-volatile memory system is disclosed. The non-volatile memory system may include a non-volatile memory having a plurality of sub-drives, as well as a controller in communication with the plurality of sub-drives. The controller may be configured to sort data associated with a host write command, as the data is received from a host, into one of the plurality of sub-drives based on a determined data temperature of the data associated with the host write command. The controller may be further configured to sort data, already stored in a first sub-drive of the plurality of sub-drives, into a different sub-drive of the plurality of sub-drives by logically remapping a portion of the data already stored in the first sub-drive, without rewriting the portion of the data into a different physical location, independently of any garbage collection operation in the first sub-drive.
In a further aspect, a method is disclosed for managing data in a non-volatile memory system having a plurality of sub-drives and a controller in communication with the sub-drives. The method may include the controller receiving a host data write at the memory system and directing the host data write to one of the plurality of sub-drives based on a first sorting technique. The method may also include the controller determining whether an amount of valid data in one of the plurality of sub-drives exceeds a predetermined amount of a logical address space for the one of the plurality of sub-drives. When the amount of valid data in the one of the plurality of sub-drives exceeds the predetermined amount, the method includes re-mapping a superblock from the one of the plurality of sub-drives to another of the plurality of sub-drives, without rewriting any data from the superblock to another physical location, based on a current data temperature of the superblock.
Separately or in combination with the above-noted method, a method for initially sorting data in a non-volatile memory system is described. The method may be executed in a non-volatile memory system with a plurality of sub-drives each associated with a different data temperature range. The controller of the system may maintain a list of most recent host data writes for one of the plurality of sub-drives. The method may include comparing the logical addresses of data in a received host write command to logical addresses of data in the list of most recent host data writes for the one of the plurality of sub-drives. When a logical address of data in the received host data write is present in the list, the method includes automatically routing the data in the received host write command to a different one of the plurality of sub-drives associated with a hotter data temperature range than the one of the plurality of sub-drives.
Referring now to
The controller 102 (which may be a flash memory controller) can take the form of processing circuitry, one or more microprocessors or processors (also referred to herein as central processing units (CPUs)), and a computer-readable medium that stores computer-readable program code (e.g., software or firmware) executable by the (micro)processors, logic gates, switches, an application specific integrated circuit (ASIC), a programmable logic controller, and an embedded microcontroller, for example. The controller 102 can be configured with hardware and/or firmware to perform the various functions described below and shown in the flow diagrams. Also, some of the components shown as being internal to the controller can also be stored external to the controller, and other components can be used. Additionally, the phrase “operatively in communication with” could mean directly in communication with or indirectly (wired or wireless) in communication with through one or more components, which may or may not be shown or described herein.
As used herein, a flash memory controller is a device that manages data stored on flash memory and communicates with a host, such as a computer or electronic device. A flash memory controller can have various functionality in addition to the specific functionality described herein. For example, the flash memory controller can format the flash memory to ensure the memory is operating properly, map out bad flash memory cells, and allocate spare cells to be substituted for future failed cells. Some part of the spare cells can be used to hold firmware to operate the flash memory controller and implement other features. In operation, when a host needs to read data from or write data to the flash memory, it will communicate with the flash memory controller. If the host provides a logical address to which data is to be read/written, the flash memory controller can convert the logical address received from the host to a physical address in the flash memory. The flash memory controller can also perform various memory management functions, such as, but not limited to, wear leveling (distributing writes to avoid wearing out specific blocks of memory that would otherwise be repeatedly written to) and garbage collection (after a block is full, moving only the valid pages of data to a new block, so the full block can be erased and reused).
Non-volatile memory die 104 may include any suitable non-volatile storage medium, including NAND flash memory cells and/or NOR flash memory cells. The memory cells can take the form of solid-state (e.g., flash) memory cells and can be one-time programmable, few-time programmable, or many-time programmable. The memory cells can also be single-level cells (SLC), multiple-level cells (MLC), triple-level cells (TLC), or use other memory cell level technologies, now known or later developed. Also, the memory cells can be fabricated in a two-dimensional or three-dimensional fashion.
The interface between controller 102 and non-volatile memory die 104 may be any suitable flash interface, such as Toggle Mode 200, 400, or 800. In one embodiment, memory system 100 may be a card based system, such as a secure digital (SD) or a micro secure digital (micro-SD) card. In an alternate embodiment, memory system 100 may be part of an embedded memory system.
Although in the example illustrated in
Modules of the controller 102 may include a sub-drive data routing module 112 present on the die of the controller 102. As described below, the sub-drive data routing module 112 may provide functionality for routing data from a host only to a particular portion of non-volatile memory 104 and for moving data at predetermined times between various portions of the non-volatile memory 104 based on a frequency of host activity regarding the data. The sub-drive data routing module 112 of the controller 102 may accomplish this by tracking activity (e.g. the number of host writes or the number of host reads) to individual logical block addresses (LBAs) in predefined sections of contiguous LBAs, referred to herein as LBA blocks, in the logical address space. The sub-drive data routing module 112 may then assign an average activity count, also referred to herein as temperature, to all the LBAs included in that particular LBA block and, upon initiation of a garbage collection operation in a sub-drive, move data associated with a particular LBA to a physical block in the same or another sub-drive based on the temperature associated with that LBA. The sub-drive data routing module 112 may also manage sub-drives differently in the NVM system 100 such that only one sub-drive includes an open host write block, thus is the only sub-drive accepting host data from the host. Also, all other sub-drives, except for the single sub-drive that accepts host data, include open relocation blocks for accepting data relocated from a garbage collection operation. In other words, in one implementation all data from the host must always first go to the single sub-drive dedicated to receive host data and all other sub-drives only receive relocated data from each other or the single dedicated sub-drive (referred to herein as a staging sub-drive).
A buffer manager/bus controller 114 manages buffers in random access memory (RAM) 116 and controls the internal bus arbitration of controller 102. A read only memory (ROM) 118 stores system boot code. Although illustrated in
The RAM 116 in the NVM system 100, whether outside the controller 102, inside the controller or present both outside and inside the controller 102, may contain a number of items, including a copy of one or more pieces of the logical-to-physical mapping tables for the NVM system 100. The RAM 116 may contain block activity data 117 that is used to track the temperature of data in the various sub-drives. The block activity data may be maintained at different levels of granularity, for example on a superblock level, in the form of individual LBA counters (or counters for the average temperature/activity for larger groups of LBAs) and so on. The RAM 116 may also include a free block list 121 indicating currently unused physical superblocks available for use in any of the sub-drives of the non-volatile memory 104.
Front end module 108 includes a host interface 120 and a physical layer interface (PHY) 122 that provide the electrical interface with the host or next level storage controller. The choice of the type of host interface 120 can depend on the type of memory being used. Examples of host interfaces 120 include, but are not limited to, SATA, SATA Express, SAS, Fibre Channel, USB, PCIe, and NVMe. The host interface 120 typically facilitates transfer for data, control signals, and timing signals.
Back end module 110 includes an error correction controller (ECC) engine 124 that encodes the data bytes received from the host, and decodes and error corrects the data bytes read from the non-volatile memory. A command sequencer 126 generates command sequences, such as program and erase command sequences, to be transmitted to non-volatile memory die 104. A RAID (Redundant Array of Independent Drives) module 128 manages generation of RAID parity and recovery of failed data. The RAID parity may be used as an additional level of integrity protection for the data being written into the NVM system 100. In some cases, the RAID module 128 may be a part of the ECC engine 124. A memory interface 130 provides the command sequences to non-volatile memory die 104 and receives status information from non-volatile memory die 104. In one embodiment, memory interface 130 may be a double data rate (DDR) interface, such as a Toggle Mode 200, 400, or 800 interface. A flash control layer 132 controls the overall operation of back end module 110.
Additional components of NVM system 100 illustrated in
In one implementation, an individual data latch may be a circuit that has two stable states and can store 1 bit of data, such as a set/reset, or SR, latch constructed from NAND gates. The data latches 158 may function as a type of volatile memory that only retains data while powered on. Any of a number of known types of data latch circuits may be used for the data latches in each set of data latches 158. Each non-volatile memory die 104 may have its own sets of data latches 158 and a non-volatile memory array 142. Peripheral circuitry 141 includes a state machine 152 that provides status information to controller 102. Peripheral circuitry 141 may also include additional input/output circuitry that may be used by the controller 102 to transfer data to and from the latches 158, as well as an array of sense modules operating in parallel to sense the current in each non-volatile memory cell of a page of memory cells in the non-volatile memory array 142. Each sense module may include a sense amplifier to detect whether a conduction current of a memory cell in communication with a respective sense module is above or below a reference level.
The non-volatile flash memory array 142 in the non-volatile memory 104 may be arranged in blocks of memory cells. A block of memory cells is the unit of erase, i.e., the smallest number of memory cells that are physically erasable together. For increased parallelism, however, the blocks may be operated in larger metablock units. One block from each of at least two planes of memory cells may be logically linked together to form a metablock. Referring to
The individual blocks are in turn divided for operational purposes into pages of memory cells, as illustrated in
The term superblock may be used interchangeably with the term metablock herein. A superblock, as described in more detail below with reference to
Referring to
In
As noted previously and described in greater detail below, the temperature of a given LBA refers to the frequency of host activity regarding the LBA associated with the data where, in one implementation, host activity refers only to host write operations. Although in other implementations the host read operations may also be considered as host activity, the discussion herein focuses on write activity as tracking and utilizing write activity may provide desired write amplification benefits. A simplified version of tracking data temperature is described herein, where temperature may be tracked on a superblock granularity basis such that all LBAs in a given superblock are given the same temperature. Other granularities of assigning and tracking temperature are also contemplated.
As shown in
Each of the sub-drives 604, 606, 608 is a collection of superblocks that are managed together. Also, each of the sub-drives 604, 606, 608 may exist in separate non-volatile memory die 104, the same non-volatile memory die, or each straddle multiple non-volatile memory die 104 in the NVM system 100. Each sub-drive may include only one type of memory cell, such as SLC or MLC, or multiple types of memory cells.
The initial routing of data from the host to the sub-drives and between sub-drives is managed by the sub-drive data routing module 112 in the controller 102. With respect to initial host writes to a sub-drive, any of a number of hot/cold sorting functions 602 may be implemented. In one implementation, the sorting function 602 of the sub-drive data routing module 112 is configured such that all data coming from a host write into the non-volatile memory 600 is only sent initially to an open host write superblock 603 in the sub-drive that the LBAs of the incoming host write were last written to. If the superblock of incoming data for the host write includes updated data for LBAs that are currently in superblocks in different sub-drives, then the incoming superblock of host data may be written to the highest temperature sub-drive associated with the incoming LBAs. In another implementation, the sorting function 602 may cause the controller 102 to direct incoming host writes to a next hottest sub-drive. For example, if the LBAs in the superblock of the incoming host write data were previously associated with a superblock 500 in the cold sub-drive 608, the incoming host write would be directed to the intermediate sub-drive 606.
In other implementations of the hot/cold sorting algorithm 602, all incoming host write data may be written to a dedicated staging sub-drive (not shown) that is not associated with a temperature and from which the host write data may be remapped into one of the temperature-associated sub-drives 604, 606, 608. It is also contemplated that all incoming host data may initially be written to the hot sub-drive 604. In yet another implementation, the hot/cold sorting function 602 for incoming host writes may simply use a round robin technique that takes turns directing sequential superblock host writes to a different sub-drive 604, 606, 608 each time. As discussed in greater detail herein, another initial host superblock write routing technique is contemplated, which looks at recent write density of the LBAs in the superblock, that may react more quickly to changes in the write activity associated with LBAs stored in a particular superblock. Any of a number of other initial host write routing algorithms are also contemplated.
In one implementation, after data is routed to a desired sub-drive during the initial host write, valid data from a selected source superblock is only re-written during a garbage collection operation to an open relocation superblock 605 into the same sub-drive as the source superblock for that valid data.
Data flow possibilities into and between the sub-drives 604, 606, 608 of the non-volatile memory 600 are shown in
In contrast to the solid lines of initial host write paths 612 and garbage collection paths 616, possible re-mapping paths 614 of superblocks from one sub-drive to another are shown as dotted lines. The re-mapping of superblocks between sub-drives is accomplished logically, separately from any garbage collection activity, such that no data is rewritten to different physical locations. This allows the NVM system 100 to use separate criteria for re-mapping superblocks to a different temperature sub-drive and for selecting a source block for garbage collection. For example, if temperature sorting was combined with garbage collection then the sorting may be slowed down by the fact that the coldest block in a higher temperature sub-drive that may need to be demoted from that higher temperature sub-drive to a colder temperature sub-drive, may not be the same block that is most efficient for purposes of garbage collection from the sub-drive. So, if the superblock with the least amount of valid data (typically selected for a garbage collection operation) does not coincide with the coldest superblock in that sub-drive, then the hot/cold data sorting by garbage collection may be less accurate. The re-mapping of superblocks may be triggered by one of the sub-drives overflowing its allotted logical address space and the controller 102 may select a coldest block for re-mapping to the next colder sub-drive (or hottest block for re-mapping to the next hotter drive, for example if the cold sub-drive 608 is too full) and re-map that superblock into the other sub-drive.
Referring to
Also illustrated in
The boundaries of the temperature ranges assigned to the sub-drives, such as the hot, medium and cold sub-drives in the examples of
Any of a number of different data temperature tracking mechanisms may be used to monitor the temperature of data written into the sub-drives of the NVM system 100. Two different data temperature tracking mechanisms that track data temperature at the superblock level are illustrated in
In one implementation, the controller 102, via the sub-drive data routing module 112, keeps track of all host write transactions for every LBA in every superblock for the life of the NVM system 100. The sub-drive data routing module 112 may store this information in the block activity data 117 maintained in RAM 116. Thus, the predetermined amount of time within which the number of host writes to an LBA are tracked may be the entire life of the NVM system 100 and a workload for the sub-drives may be based on the percentage of the total host writes to LBAs in each particular sub-drive. In other implementations, in order to provide a more useful indication of changes in host write density to LBAs and the associated sub-drives containing those LBAs, the predetermined amount of time may refer to a fixed-size window of time that extends back a limited amount of time to the present, wherein only the most recent host writes are included in the workload calculation and earlier LBA hits are removed from the respective counters.
A visual representation of this sliding window of time within which the host write activity at LBAs is shown in
A method of utilizing the NVM system 100 with sub-drives and data structures described above is illustrated in
Periodically, for example after every host write to the NVM system 100, the controller 102 may determine whether a garbage collection operation is needed for any of the sub-drives (at 1006). One suitable garbage collection trigger may be the number of free blocks in the free block pool shared by the sub-drives in the non-volatile memory 104 falling below a predetermined minimum value. If the controller 102 detects that too few free blocks are available based on the free block list 121, then a garbage collection operation may be initiated. The number of free blocks is just one example of a garbage collection trigger and different or additional garbage collection triggers are contemplated.
Once triggered, the sub-drive in the non-volatile memory 104 is selected from which a source superblock will be garbage collected (at 1008). One sub-drive selection process, based on selecting the sub-drive that currently has an amount of over-provisioning (OP) that is the most over its calculated targeted OP amount, is described in greater detail below. Once the sub-drive is selected, the a source superblock is selected from that sub-drive (1010), for example the superblock 500 having the least amount of valid data. Valid data from the selected source superblock is then copied to a destination (relocation) superblock in the same sub-drive (at 1012). As noted previously, the valid data copied from a source superblock is only copied into another superblock in the same sub-drive and no data crosses sub-drive boundaries during garbage collection. The sorting of previously written data from one sub-drive to a different sub-drive based on data temperature happens separately in a logical re-mapping process described below. After all valid data pieces from the selected source superblock 500 have been relocated, then the selected source superblock may be added to the free block list 121 maintained in RAM 116 (at 1014). Superblocks 500 in the free block list 121 may be later used in any of the sub-drives as needed.
Referring again to
In one implementation, the superblock selected for logical re-mapping from the sub-drive identified as being over its allotted logical address amount may be the coldest superblock in that sub-drive and may be re-mapped to the next coldest sub-drive as illustrated in
Referring again to the garbage collection decision path of
The target overprovisioning thresholds for each sub-drive in the non-volatile memory, as described in greater detail below, are determined based on the current logical capacity occupied by valid data in each sub-drive and the current write traffic (also referred to herein as workload) in each sub-drive. The NVM system 100, through these variable target overprovisioning settings, takes advantage of the pattern of data movement between sub-drives driven by the data temperature boundaries assigned the sub-drives to help avoid write amplification for “hotter” (higher update frequency) data. For example, the NVM system 100 may be configured at manufacture to include a predetermined amount of minimum physical capacity overprovisioning for the system as a whole and then divide up the logical capacity as desired among the sub-drives where a sub-drive associated with a data type having a higher likelihood of update is allocated a larger capacity overprovisioning than a sub-drive associated with a data type having a lower frequency of update.
As noted above, the non-volatile memory 104 of the NVM system 100, as a whole, may be manufactured with an overprovisioning amount such that there is a predetermined extra physical capacity (e.g., extra physical superblocks) greater than the predetermined total logical capacity. Each sub-drive in the NVM system 100 possesses a variable portion of the NVM system's logical capacity based on the historical flow of data into each sub-drive based on predetermined. Also, for a given window of time, such as window 904 (
Based on the current logical capacity, current workload and the total overprovisioning of the entire NVM system, the controller 102 via the sub-drive routing module 112 can then calculate a target over-provisioning to optimize the physical overcapacity for each sub-drive. Finally, by selecting a superblock for garbage collection from a sub-drive that currently is above its respective calculated target overprovisioning amount, the overprovisioning can be adjusted. Thus, the overprovisioning of each sub-drive may be determined based on the workflow (i.e., access density in terms of the number of host writes) and selection of a superblock for a garbage collection operation may be made based on the relative overprovisioning status of each sub-drive.
Any of a number of target overprovisioning formulas and techniques may be used to determine target overprovisioning for each sub-drive.
One suitable algorithm may be:
where θi is the fraction (0 to 1) of target OP that should be applied to sub-drive i; αi is the fraction (0 to 1) of total NVM system valid logical space in sub-drive i; and vi is the fraction (0 to 1) of total host write workload W attributed to sub-drive i. In the examples of
Additionally, a somewhat cruder, more iterative “hit or miss” method of determining desired overprovisioning for each of the sub-drives may be implemented based on an analysis of a write amplification surface generated for the NVM system. An example of a standard write amplification algorithm is
where a write amplification surface may be generated as the iterative sum of each of k sub-drives (SD) group hit rate multiplied by the write amplification for that sub-drive at a given overprovisioning level WriteAmp(OPSD
Initial Host Write with Hot Host List
A method of sorting initially written host data to a desired sub-drive based on data temperature is illustrated in
The method includes keeping a separate first in first out (FIFO) recent write list of superblocks for each sub-drive other than the hottest sub-drive. Referring to the three sub-drive non-volatile memory 104 version of
The initial host write sort may reduce the endurance penalty due to inaccuracies of hot/cold data copies while not requiring a lot of system resources. It is based on indirect measurement of host write density using block-level algorithm with adjustable accuracy-speed ratio as illustrated in
Referring to
Eventually, the new hot data being remapped, for example from SD1 to SD0, triggers remapping of one of the oldest block from SD0 to SD1 (see dashed lines between SD0 and SD1 in
If the workload pattern changes so that host write ‘hot spot’ moves to another set of LBAs, the data remapping from sub-drive to sub-drive will happen as soon as the host writes the hot data twice, in cases where the second host write to the LBA happens within the fixed number of writes tracked in HHL1 or HHL2, while cold data will get re-mapped to a colder SD. In other embodiments, a hot host list (HHL1 and HHL2) may also include a super-block written by garbage collection and not just host writes if there are not enough blocks written by the host or/and there are GC blocks containing relatively recently written data (for example, when the superblocks on the HHLs are garbage collected and thus the garbage collection superblocks contain hot data). Ordering within the list would maintain the order to keep the block in approximate order of time since written by the host. For example, in one implementation where the HHLs include garbage collection activity as well as host writes, the HHL may be the left-most entries (e.g. most recent 4 entries) of the data age lists 800 or 802 of
Referring to
The left-most points of the plot 1300 in
Although the initial host write sorting method of
In the present application, semiconductor memory devices such as those described in the present application may include volatile memory devices, such as dynamic random access memory (“DRAM”) or static random access memory (“SRAM”) devices, non-volatile memory devices, such as resistive random access memory (“ReRAM”), electrically erasable programmable read only memory (“EEPROM”), flash memory (which can also be considered a subset of EEPROM), ferroelectric random access memory (“FRAM”), and magnetoresistive random access memory (“MRAM”), and other semiconductor elements capable of storing information. Each type of memory device may have different configurations. For example, flash memory devices may be configured in a NAND or a NOR configuration.
The memory devices can be formed from passive and/or active elements, in any combinations. By way of non-limiting example, passive semiconductor memory elements include ReRAM device elements, which in some embodiments include a resistivity switching storage element, such as an anti-fuse, phase change material, etc., and optionally a steering element, such as a diode, etc. Further by way of non-limiting example, active semiconductor memory elements include EEPROM and flash memory device elements, which in some embodiments include elements containing a charge storage region, such as a floating gate, conductive nanoparticles, or a charge storage dielectric material.
Multiple memory elements may be configured so that they are connected in series or so that each element is individually accessible. By way of non-limiting example, flash memory devices in a NAND configuration (NAND memory) typically contain memory elements connected in series. A NAND memory array may be configured so that the array is composed of multiple strings of memory in which a string is composed of multiple memory elements sharing a single bit line and accessed as a group. Alternatively, memory elements may be configured so that each element is individually accessible, e.g., a NOR memory array. NAND and NOR memory configurations are exemplary, and memory elements may be otherwise configured.
The semiconductor memory elements located within and/or over a substrate may be arranged in two or three dimensions, such as a two-dimensional memory structure or a three-dimensional memory structure.
In a two dimensional memory structure, the semiconductor memory elements are arranged in a single plane or a single memory device level. Typically, in a two-dimensional memory structure, memory elements are arranged in a plane (e.g., in an x-z direction plane) which extends substantially parallel to a major surface of a substrate that supports the memory elements. The substrate may be a wafer over or in which the layer of the memory elements are formed or it may be a carrier substrate which is attached to the memory elements after they are formed. As a non-limiting example, the substrate may include a semiconductor such as silicon.
The memory elements may be arranged in the single memory device level in an ordered array, such as in a plurality of rows and/or columns. However, the memory elements may be arrayed in non-regular or non-orthogonal configurations. The memory elements may each have two or more electrodes or contact lines, such as bit lines and word lines.
A three-dimensional memory array is arranged so that memory elements occupy multiple planes or multiple memory device levels, thereby forming a structure in three dimensions (i.e., in the x, y and z directions, where the y direction is substantially perpendicular and the x and z directions are substantially parallel to the major surface of the substrate).
As a non-limiting example, a three-dimensional memory structure may be vertically arranged as a stack of multiple two dimensional memory device levels. As another non-limiting example, a three-dimensional memory array may be arranged as multiple vertical columns (e.g., columns extending substantially perpendicular to the major surface of the substrate, i.e., in the y direction) with each column having multiple memory elements in each column. The columns may be arranged in a two dimensional configuration, e.g., in an x-z plane, resulting in a three dimensional arrangement of memory elements with elements on multiple vertically stacked memory planes. Other configurations of memory elements in three dimensions can also constitute a three-dimensional memory array.
By way of non-limiting example, in a three dimensional NAND memory array, the memory elements may be coupled together to form a NAND string within a single horizontal (e.g., x-z) memory device levels. Alternatively, the memory elements may be coupled together to form a vertical NAND string that traverses across multiple horizontal memory device levels. Other three dimensional configurations can be envisioned wherein some NAND strings contain memory elements in a single memory level while other strings contain memory elements which span through multiple memory levels. Three dimensional memory arrays may also be designed in a NOR configuration and in a ReRAM configuration.
Typically, in a monolithic three dimensional memory array, one or more memory device levels are formed above a single substrate. Optionally, the monolithic three-dimensional memory array may also have one or more memory layers at least partially within the single substrate. As a non-limiting example, the substrate may include a semiconductor such as silicon. In a monolithic three-dimensional array, the layers constituting each memory device level of the array are typically formed on the layers of the underlying memory device levels of the array. However, layers of adjacent memory device levels of a monolithic three-dimensional memory array may be shared or have intervening layers between memory device levels.
Then again, two-dimensional arrays may be formed separately and then packaged together to form a non-monolithic memory device having multiple layers of memory. For example, non-monolithic stacked memories can be constructed by forming memory levels on separate substrates and then stacking the memory levels atop each other. The substrates may be thinned or removed from the memory device levels before stacking, but as the memory device levels are initially formed over separate substrates, the resulting memory arrays are not monolithic three dimensional memory arrays. Further, multiple two dimensional memory arrays or three dimensional memory arrays (monolithic or non-monolithic) may be formed on separate chips and then packaged together to form a stacked-chip memory device.
Associated circuitry is typically required for operation of the memory elements and for communication with the memory elements. As non-limiting examples, memory devices may have circuitry used for controlling and driving memory elements to accomplish functions such as programming and reading. This associated circuitry may be on the same substrate as the memory elements and/or on a separate substrate. For example, a controller for memory read-write operations may be located on a separate controller chip and/or on the same substrate as the memory elements.
One of skill in the art will recognize that this invention is not limited to the two-dimensional and three-dimensional exemplary structures described but cover all relevant memory structures within the spirit and scope of the invention as described herein and as understood by one of skill in the art.
Methods and systems have been disclosed for managing received data in a multiple sub-drive non-volatile memory system. The data may be managed through a direct write process where data is routed through an initial sorting algorithm to sub-drives in non-volatile memory and subsequent sorting of data and garbage collection operations in the non-volatile memory are performed independently of one another. Separately, or in combination with the independent garbage collection and logical sorting of superblocks of data, a method of initially sorting host write data into sub-drives based on a write density temperature measurement is disclosed. The systems and methods may permit reduction of write amplification, and thus overall improvement of IOPS, for a non-volatile memory system.
It is intended that the foregoing detailed description be understood as an illustration of selected forms that the invention can take and not as a definition of the invention. It is only the following claims, including all equivalents, that are intended to define the scope of the claimed invention. Finally, it should be noted that any aspect of any of the preferred embodiments described herein can be used alone or in combination with one another.
Gorobets, Sergey Anatolievich, Parker, Liam Michael
Patent | Priority | Assignee | Title |
Patent | Priority | Assignee | Title |
8239639, | Jun 08 2007 | SanDisk Technologies LLC | Method and apparatus for providing data type and host file information to a mass storage system |
8261009, | Dec 30 2008 | Sandisk IL Ltd. | Method and apparatus for retroactive adaptation of data location |
9021231, | Sep 02 2011 | SanDisk Technologies LLC | Storage control system with write amplification control mechanism and method of operation thereof |
9176864, | May 17 2011 | SanDisk Technologies LLC | Non-volatile memory and method having block management with hot/cold data sorting |
20120059978, | |||
20120297122, | |||
20130061019, | |||
20170285948, | |||
20180190329, | |||
20180357159, |
Executed on | Assignor | Assignee | Conveyance | Frame | Reel | Doc |
Jun 12 2017 | GOROBETS, SERGEY ANATOLIEVICH | Western Digital Technologies, INC | ASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS | 045561 | /0130 | |
Jun 12 2017 | PARKER, LIAM MICHAEL | Western Digital Technologies, INC | ASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS | 045561 | /0130 | |
Apr 16 2018 | Western Digital Technologies, Inc. | (assignment on the face of the patent) | / | |||
Aug 01 2019 | GOROBETS, SERGEY ANATOLIEVICH | Western Digital Technologies, INC | ASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS | 049936 | /0509 | |
Aug 01 2019 | PARKER, LIAM MICHAEL | Western Digital Technologies, INC | ASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS | 049936 | /0509 | |
Jan 13 2020 | Western Digital Technologies, INC | JPMORGAN CHASE BANK, N A , AS AGENT | SECURITY INTEREST SEE DOCUMENT FOR DETAILS | 052915 | /0566 | |
Feb 03 2022 | JPMORGAN CHASE BANK, N A | Western Digital Technologies, INC | RELEASE OF SECURITY INTEREST AT REEL 052915 FRAME 0566 | 059127 | /0001 | |
Aug 18 2023 | Western Digital Technologies, INC | JPMORGAN CHASE BANK, N A | PATENT COLLATERAL AGREEMENT - A&R LOAN AGREEMENT | 064715 | /0001 | |
Aug 18 2023 | Western Digital Technologies, INC | JPMORGAN CHASE BANK, N A | PATENT COLLATERAL AGREEMENT - DDTL LOAN AGREEMENT | 067045 | /0156 | |
May 03 2024 | Western Digital Technologies, INC | SanDisk Technologies, Inc | ASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS | 067567 | /0682 | |
Jun 21 2024 | SanDisk Technologies, Inc | SanDisk Technologies, Inc | CHANGE OF NAME SEE DOCUMENT FOR DETAILS | 067982 | /0032 |
Date | Maintenance Fee Events |
Apr 16 2018 | BIG: Entity status set to Undiscounted (note the period is included in the code). |
May 01 2023 | REM: Maintenance Fee Reminder Mailed. |
Jun 07 2023 | M1551: Payment of Maintenance Fee, 4th Year, Large Entity. |
Jun 07 2023 | M1554: Surcharge for Late Payment, Large Entity. |
Date | Maintenance Schedule |
Sep 10 2022 | 4 years fee payment window open |
Mar 10 2023 | 6 months grace period start (w surcharge) |
Sep 10 2023 | patent expiry (for year 4) |
Sep 10 2025 | 2 years to revive unintentionally abandoned end. (for year 4) |
Sep 10 2026 | 8 years fee payment window open |
Mar 10 2027 | 6 months grace period start (w surcharge) |
Sep 10 2027 | patent expiry (for year 8) |
Sep 10 2029 | 2 years to revive unintentionally abandoned end. (for year 8) |
Sep 10 2030 | 12 years fee payment window open |
Mar 10 2031 | 6 months grace period start (w surcharge) |
Sep 10 2031 | patent expiry (for year 12) |
Sep 10 2033 | 2 years to revive unintentionally abandoned end. (for year 12) |